Reinterpreting Confidence Effect Experiment
Eliaz and Schotter (2007) had players make a risky choice between Box A and Box B–just one box had a cash reward. In both “states of nature” H and L, the reward was more likely to be in Box A, but the probability was higher in state H. Subjects could pay a fee to learn what state of nature was operative. Despite the fact that picking Box A was the dominant decision no matter what state of nature was operative, most subjects (about 75%) paid the fee to learn the state of nature. Eliaz and Schotter explain this irrational result as the consequence of a “confidence effect” — that humans gain intrinsic utility from being more confident in their decisions.
I dispute this explanation. A better explanation, in my view, is that, by default, we follow a social norm that more information is better in economic decisions. This social norm is unquestionably adaptive in real-life economic decision-making; this useful rule of them gets transferred to laboratory decisions by default. If the game is played once or just a few times, the social norm will be followed. After repeating many times, though, the subjects would wise up to the fact that paying the fee is a waste.
One could determine whether the confidence effect or the social norm is the better explanation by replicating the experiment with subjects repeating it for many periods. Eliaz and Schotter’s confidence effect would be identical in all periods; if they are right, then fee payment should remain steady. If I’m right, then fee-paying should converge downward over time toward the optimum.